Mo Bavarian(@mobav0):Without models communicating well their decision, the bond between human and AI will get weaker. Better writing is crucial for alignment. Still, better wiring doesn’t address everything. It for example doesn’t have much bearing to clarity of internal thought, i.e. chain of thought, (which are often are not and should not be graded) or internal activations of the model. But it’s still a great step.

Mo Bavarian

@mobav0

Scaling up RL at OpenAI 🍓 Optimization and long context research before. Math PhD, MIT.

参加 October 2016

1.1K フォロー中 22.2K ファン

Mo Bavarian@mobav0

2026.09.24 23:24

Without models communicating well their decision, the bond between human and AI will get weaker. Better writing is crucial for alignment. Still, better wiring doesn’t address everything. It for example doesn’t have much bearing to clarity of internal thought, i.e. chain of thought, (which are often are not and should not be graded) or internal activations of the model. But it’s still a great step.

Alek Dimitriev@tensor_rotator

2026.09.24 21:39

Fixing Claude's writing was pivotal to reducing gradual disempowerment. Lack of understanding made us all cede decision-making to Claude because we couldn't even understand what it's saying.